Incidence and prevalence of atrial fibrillation: an analysis based on 8.3 million patients

Abstract

Aims

Based on an analysis of claims-based data of 8.298 million members of two German statutory health insurance funds, the aim of this contribution is to quantify age-/gender-specific prevalence/incidence of atrial fibrillation (AF) in a German setting.

Methods and results

Patients were classified as AF prevalent, if they had received at least two outpatient diagnoses of AF (ICD10-Code I48.1-) in two different quarters of the year and/or had received at least one main AF diagnosis during inpatient treatment between 1 January 2007 and 12 December 2008. They were considered to have had new onset AF in 2008 under the following conditions; first, there was no AF diagnosis in 2007; secondly, patients had not received oral anticoagulant medication in 2007; and thirdly, patients had received either one inpatient/two outpatient diagnoses of AF in 2008. In our sample, a total of 176 891 patients had AF. AF prevalence was 2.132%. The average age of these AF patients was 73.1 years, and 55.5% (98 190 patients) were male. The incidence of AF in our sample was 4.358 cases/1000 person-years in men and 3.868 cases/1000 person-years in women.

Conclusion

A comparison of the distribution of AF prevalence/incidence in our population with that in already published studies showed that our figures were higher, especially in the age groups above 70 years. Our data show that in a large industrial nation such as Germany care provision structures are going to be challenged by a requirement to treat more AF patients in the future.

Introduction

Atrial fibrillation (AF) is the most significant common disorder of cardiac rhythm. It is associated with substantial morbidity and mortality from stroke and thromboembolism.^1–4 Over the past few decades, the prevalence of AF in at least the industrialized nations is known to have markedly increased,^1,5 yet as a rule most of the available data concerning AF prevalence has been gathered on the basis of patient selection bias. It has been common, for example, to include only older patients,^6–8 those willing to take part in medical screening programmes,⁹ visiting a doctor's office or a hospital^10,11 or having recently experienced a stroke.¹² Only some of the published studies on AF prevalence can be called population based,^2,9–23 and few of them report age- and/or gender-specific AF prevalence rates.

Similarly, representative conclusions concerning the age- and gender-specific incidence of AF remain rare, as such studies require time frame analyses based on person-years, ideally in the separate gender and age groups in use.^8,9,13,14,24 The authors of this paper are only aware of four such analyses, which draw on data from 1990 to 1993, 1993 to 2007, 2001 to 2003 and 1993 to 2005,^8,9,13,14 respectively.

Based on an analysis of claims-based data of 8.298 million members of two large German statutory health insurance funds (SHIFs) in 2008, the aim of this contribution is to update and more precisely quantify age- and gender-specific AF prevalence and incidence in a German setting. Moreover, by incorporating German population statistics, this study estimates the total number of German AF patients in 2009 and makes a prognosis of the likely prevalence and incidence of AF in Germany in 2020.

Methods

Study population

At present in Germany, more than 85% of the population receives mandatory heath-care provision, through a total of 221 SHIFs (2008 data). The SHIFs possess claims-based data on the inpatient and outpatient medical care provided to these patients, as well as the corresponding diagnoses. The data are collected primarily for accounting purposes, but over the past few years with increasing frequency also been used for research.²⁵ The present data set is derived from two of the largest funds, currently insuring 14.8% of the SHIF-insured population. One of the funds is active in all the regions of the country; the other is active in two federal states in the southeast. Data were available for the years 2007 and 2008. Our analysis included the relevant data about all the patients, who were continuously insured with one of the funds during these two years (8.298 million people).

Atrial fibrillation prevalence

Patients were classified as AF prevalent, if they had received at least two outpatient diagnoses of AF (ICD10-Code I48.1-) in two different quarters between 1 January 2007 and 31 December 2008 and/or had received at least one main AF diagnosis during inpatient treatment in these two years. Patients who had only received one outpatient diagnosis of AF were not considered definitely to be AF patients, because it was possible that suspected cases could become mixed up with confirmed cases. In the subsequent data analysis, all forms of AF were included, but atrial flutter (ICD10-Code I48.0-) was excluded.

The age- and gender-specific prevalence of AF was estimated by dividing the number of AF patients registered with the funds until 31 December 2008 by the total number of patients in the appropriate and predetermined age/gender group. The assignment to an age group was based on the age of a patient on 01 January 2008.

Atrial fibrillation incidence

A patient was considered to have had new onset AF in 2008 under three conditions; first, the patient had not had a diagnosis of AF in 2007; secondly, this patient had not received oral anticoagulant medication in 2007; and thirdly, the patient had received either one inpatient or two outpatient diagnoses of AF conducted in two different quarters of 2008.

Age-/gender-specific incidence was reported as cases per 1000 person-years. In all cases, the period of observation used as the time frame for calculating these person-years began on 1 January 2008. For each patient experiencing new onset AF in 2008, the observation period ended on the day of the first inpatient or second outpatient diagnosis in 2008 and for the other patients in the sample the period ended on 31 December 2008.

Statistical analysis and comparison with literature

After classifying the sample population by gender, we used groups formed by 5-year age intervals to estimate the prevalence and incidence of AF. Since we were also interested in providing a perspective on future trends in the prevalence and incidence of AF, we used the official German population statistics.²⁶ The data from 2009 were used as the base year. Furthermore, the official population prognosis for 2020 was used to make predictions concerning age-/gender-specific AF prevalence/incidence.

Our AF prevalence/incidence estimates were then compared to international studies on AF prevalence and incidence that have also published gender-/age-specific data. Relevant studies were identified by a Pubmed/MEDLINE research (PRISMA Figure A in the Supplementary material). To compare published AF prevalence/incidence results with our numbers, the average level of the male/female prevalence/incidence (arithmetic mean) in different age groups reported in each of these reports was calculated and used as a basis for the comparison with our data. All calculations were conducted using the Microsoft SQL Server 2008 (Microsoft Inc., USA); Microsoft Excel 2010 (Microsoft Inc.) was used for statistical analyses.

Results

Atrial fibrillation prevalence and sample characteristics

In our sample, a total of 176 891 patients had AF in 2008. Compared to the total membership of the two medical insurance funds, the prevalence of AF was 2.132%. The average age of these AF patients was 73.1 years, and 55.5% (98 190 patients) were male (Table 1; detailed numbers are available in Table B in the Supplementary material). The average CHADS₂-score, based on documented diagnoses in 2007/2008, was 2.8 (CHA₂DS₂-VASc-score: 4.2). The main co-morbidities present were hypertension (87.4%), both types of diabetes mellitus (43.3%), heart failure (42.1%), asthma bronchiale/COPD (33.8%), and vascular disease (20.4%). The five most common prescribed medications (defined as those for which a patient had received at least two prescriptions in 2008) were phenprocoumon (51.7%), metoprolol (38.7%), torasemid (31.6%), bisoprolol (30.6%), and simvastatin (30.5%).

In all age classes in our data set, men had a higher prevalence of AF than women (Table 2; P < 0.001). With the exception of the oldest group (>89 years), the prevalence of AF increased successively with age; however, because of small number of patients in the older age groups, not all prevalence differences between age groups were significant. Based on our descriptive statistics only, in each subsequent age group, the prevalence rate increased successively, reaching a peak of 15.07% in the 85–89 years age group. The oldest patients (>89 years) exhibited less AF (12.70%) than the age group below (P < 0.001). A comparison of the distribution of AF prevalence in our population with that in already published studies, of which none was specifically linked to Germany, showed that our figures seem to be higher, especially in the age groups above 70 years (Figure 1).

Atrial fibrillation incidence

The incidence of AF in our sample was 4.358 cases per 1000 person-years in men and 3.868 cases in 1000 person-years in women (Table 2; detailed numbers are available in Table B in the Supplementary material). For men, the incidence increased from under 0.015 cases per 1000 person-years in the age group <15 years to 37.258 cases in 1000 person-years in age group >89 years; again, not all differences between the age groups reached significance. The corresponding rates for women were 0.018 cases per 1000 person-years (age group <15 years) and 29.025 cases per 1000 person-years (age group >89 years). For men, the AF incidence reached its peak in the oldest age group, above 89 years (P = 0.046 in a comparison with age group 85–89 years). For women, the peak was reached in the second-oldest age group (30.371 cases in 1000 person-years). However, that difference did not reach significance in a comparison to the oldest age group (P = 0.136). In comparison with existing publications, our investigation seems to show a generally higher incidence of AF, with the exception of the oldest age group (Figure 2).

Atrial fibrillation patients in Germany

Based on our age- and gender-specific prevalence data and the official population statistics for 2009, we estimate that 1 793 277 patients in Germany were AF prevalent in that year. Our subsequent incidence calculations estimate that in 2009, 362 970 patients were diagnosed with new onset AF. Based on the expected demographic changes in the German population, the number of AF patients is likely to increase to 2 126 656 in 2020; the number of patients with diagnosed new onset AF in that year would be 426 027. In summary, and assuming a constant level of AF prevalence/incidence in the specific age/gender groups, the overall AF prevalence of AF in Germany can be expected to increase by almost 0.5% points in the next 10 years because of demographic developments (Figure 3).

Discussion

Objectives and specific features of this study

Using a large German claims-based data set, we assessed the prevalence and incidence of AF and compared these data with previous studies. The use of a database covering a large patient population of all ages, a combined prevalence/incidence calculation in specific age/gender groups, and a first-time AF prevalence/incidence calculation for Germany are the distinguishing features of this study.

This paper extends and refines what is known concerning the general pattern of AF prevalence and its incidence as established by previous research. In its treatment of the age classes, our study provides the most detailed approach to both AF prevalence and incidence available to date. In addition, it is likely that in our sample there is very little bias because the SHIFs are large enough to representatively mirror the characteristics of the entire German population.

Limitations

We acknowledge some limitations of our study. First, our analysis covers only that part of the German population that receives mandatory health-care provision through the SHIFs (85% of the population). The remaining population is mostly insured by private insurance funds. The people insured by these private insurance funds are comparatively younger and more often male.²⁷ Secondly, it is known from the literature that AF prevalence numbers based on routine diagnoses mostly underestimate the number of AF patients, because not every case of AF is diagnosed and, even more importantly, many patients with AF do not visit a doctor/a hospital regularly.² As we set the criteria for sample inclusion relatively narrowly, by requiring two outpatient AF diagnoses in two different quarters or at least one inpatient AF diagnosis during an observation period of 2 years, it is unlikely that we have included outpatient patients who were merely suspected of having AF. On this basis, we believe that our figures are likely to be conservative. Thirdly, some of the existing studies used a more limited AF definition. For example, Rietbrock et al.¹⁴ only looked at chronic AF. This may explain our comparatively high AF prevalence/incidence numbers. Fourthly, we were unable to differentiate between the different clinical forms of AF, because our data do not contain this information. Fifthly, our approach to classifying a patient as a case of AF-incidence is specific. First of all, our AF definition is based on documented diagnoses. We do not know which percentages of these diagnoses were proven by electrocardiograms and/or a cardiologists’ evaluation. Furthermore, we assumed that patients should be included if they had received two outpatient or one inpatient AF diagnoses in 2008, but no outpatient/inpatient AF diagnosis in 2007. In addition, a patient was not seen as a new onset AF case if anticoagulant drugs were prescribed in 2007. We used this criterion to control for cases of AF in 2007 that had not been properly documented, not to over-estimate the AF incidence in the following year. However, we cannot say if and to what extent the actual AF incidence has been under-estimated by this approach.

Comparison with previous studies

Our analysis shows comparatively high AF prevalence and incidence rates. Nevertheless, our results are in line with the established trend towards increases in both AF prevalence and incidence that have been described in the literature over the past few decades.^11,28 A single but recent analysis of US Medicare beneficiaries (aged 65 years or older) has reported even higher AF incidences than we have (8; Figure 2). However, this may be because of its researchers using a different definition of new onset AF. Whereas we excluded all the patients who had received neither a diagnosis of AF nor oral anticoagulant medication in the previous year, Piccini et al.⁸ defined patients as incident if they had not received an AF diagnosis in the previous two years, and did not use medication as a study criterion. Another recent contribution²⁸ has reported, at least for women, an increase in the prevalence/incidence of AF between 1991 and 2008. Owing to the fact that our data collection were limited to two successive years, (2007/2008), it is incapable of being used to make reliable estimates of the development of AF prevalence/incidence over a longer time period; however, our comparatively high AF prevalence/incidence rates do not in themselves contradict a supposition that the AF prevalence/incidence in Germany have also increased over the past few decades.

Our analysis shows that in the age group >89 years, there was a lower AF prevalence and incidence compared to the adjacent age group. To date, this age group has hardly received separate consideration in the literature. Only Rietbrock et al.¹⁴ describe a similar pattern.

Conclusions

Our data indicate that AF prevalence/incidence depends on the age and gender of the patient, and that AF is a disease associated with a large number of co-morbidities. As these co-morbidities can already be observed in AF-incident patients, AF itself is not as a rule the first chronic disease these people display. It is impossible for this study to determine the extent to what the pre-existing co-morbidities might have contributed to the development of AF.

We conclude that in a large industrial nation such as Germany, the evaluation of claims-based data, in combination with population statistics, shows that care provision structures are going to be challenged by a requirement to treat many more AF patients in the future. This finding adds important and more recent information to the pre-existing reports, which contain data that originated as long as 21 years ago. Furthermore, the present study, in conjunction with other similar analyses,^15,16,28 underlines the importance of AF as a present and future health-care burden. Since many of the potential risk factors that pre-dispose to AF are modifiable, there should be an increased focus on preventing and/or reducing these risk factors, for example, better control of blood pressure, education on diet, and exercise to avoid obesity, development of hypertension, diabetes, or vascular disease, etc. and optimal medical therapy where risk factors or AF are already present.

Supplementary material

Supplementary material is available at Europace online.

Funding

The study was funded by Boehringer Ingelheim Pharma GmbH, Germany.

Acknowledgements

The authors would like to thank Mrs Dallas Reese for her efforts in the preparation of the manuscript. Moreover, they would like to thank two anonymous reviewers for their very helpful comments.

Conflict of interest: Matthias Pfannkuche is employed by Boehringer Ingelheim Pharma (Germany). Thomas Wilke, Günter Breithardt, and Rupert Bauersachs have acted as consultants for Boehringer Ingelheim Pharma, Bayer, and Bristol Myers Squibb.

References